A massively parallel multi-level approach to a domain decomposition method for the optical flow estimation with varying illumination
This work addresses computational efficiency in optical flow estimation for computer vision applications, presenting an incremental improvement through parallelization.
The authors tackled the optical flow estimation problem under varying illumination by developing a variational method with adaptive regularization and a multi-level parallel domain decomposition approach, achieving efficient computation for high-resolution images.
We consider a variational method to solve the optical flow problem with varying illumination. We apply an adaptive control of the regularization parameter which allows us to preserve the edges and fine features of the computed flow. To reduce the complexity of the estimation for high resolution images and the time of computations, we implement a multi-level parallel approach based on the domain decomposition with the Schwarz overlapping method. The second level of parallelism uses the massively parallel solver MUMPS. We perform some numerical simulations to show the efficiency of our approach and to validate it on classical and real-world image sequences.